Focus on What Matters: Two-Stage ROI-Aware Refinement for Anatomy-Preserving Fetal Ultrasound Reconstruction

2026-04-26

AI Summary

This work addresses the challenge that global metrics in fetal ultrasound reconstruction often fail to accurately assess clinically critical small regions—such as nuchal translucency—and are further compromised by multi-center domain shifts. The authors propose a parameter-free, ROI-aware two-stage reconstruction framework: the first stage optimizes global latent codes using MS-SSIM, while the second stage refines anatomically relevant regions through a combination of L1 and normalized Sobel edge constraints, with gradient magnitude automatically calibrating multi-task loss weights. The method significantly enhances key-region accuracy while preserving anatomical integrity, achieving a 6.43% reduction in ROI MAE, a 4.90% decrease in edge MAE, and a 0.29 dB PSNR gain in leave-one-hospital-out evaluation. Moreover, the latent space effectively obscures hospital-specific signatures, yielding an OOD detection AUROC of 0.9956, demonstrating strong cross-center generalization and plug-and-play capability.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Calibration & Uncertainty Quantification
• Natural Language Processing: Safety and Robustness

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Measurement-critical ultrasound tasks often depend on a small anatomical region, making global reconstruction metrics an unreliable proxy for clinical fidelity. We propose an ROI-aware representation learning framework and instantiate it for first-trimester nuchal translucency (NT) screening under multi-hospital domain shift. A two-phase convolutional autoencoder (CAE) first learns a globally faithful 128-D latent code via MS-SSIM, then refines the NT ROI using intensity (L1) and normalized Sobel-edge constraints. To combine these heterogeneous objectives without manual tuning, we initialize loss weights via gradient-based calibration from per-term gradient magnitudes. Under strict hospital-wise evaluation with one hospital held out, ROI refinement improves both global and measurement-relevant quality: on the standard dev split it increases PSNR by +0.27 dB (val) and +0.29 dB (held-out test), reduces ROI MAE by 8.87% (val) and 6.43% (held-out test), and reduces ROI Edge-MAE by 11.10% on source hospitals and 4.90% on the unseen hospital. Beyond reconstruction, frozen-latent probes provide additional evidence of generalization: hospital provenance becomes less confidently predictable on the unseen site (0.556 to 0.541 max-softmax; 0.684 to 0.688 entropy) while OOD detection remains strong across site-held-out protocols (Mahalanobis AUROC up to 0.9956, with modest KNN gains in challenging splits). The same ROI-aware refinement principle is anatomy-agnostic and can be adopted for other fetal biometry targets (e.g., crown-rump length (CRL), nasal bone (NB)) and broader medical imaging settings where small ROIs dominate clinical decisions.

Problem

• fetal ultrasound reconstruction
• region of interest (ROI)
• anatomy-preserving
• domain shift
• clinical fidelity

Innovation

• ROI-aware reconstruction
• two-stage refinement
• gradient-based loss calibration
• anatomy-preserving ultrasound
• domain generalization